debugger
Systematic debugging skill. 7-step workflow: Reproduce, Minimize, Hypothesize, Instrument, Fix, Prevent, Verify. Activate when troubleshooting errors.
SKILL.md
| Name | debugger |
| Description | Systematic debugging skill. 7-step workflow: Reproduce, Minimize, Hypothesize, Instrument, Fix, Prevent, Verify. Activate when troubleshooting errors. |
name: debugger description: Systematic debugging skill. 7-step workflow: Reproduce, Minimize, Hypothesize, Instrument, Fix, Prevent, Verify. Activate when troubleshooting errors. version: 1.0.0
Debugger 🔍
Systematic problem-solving for runtime errors, failing tests, and unexpected behavior.
[!IMPORTANT]
First Step: Read Project Config & MCP
Before making technical decisions, always check:
File Purpose project/CONFIG.yamlStack versions, modules, architecture mcp.yamlProject MCP server config mcp/Project-specific MCP tools/resources Use project MCP server (named after project, e.g.
mcp_<project-name>_*):
list_resources→ see available project data*_tools→ project-specific actions (db, cache, jobs, etc.)Use
mcp_context7for library docs:
- Check
mcp.yaml → context7.default_librariesfor pre-configured libs- Example:
libraryId: /nuxt/nuxt, query: "Nuxt 4 composables"
Language Requirements
All skill files must be in English. See LANGUAGE.md.
When to Activate
- Runtime errors, crashes
- Failing tests
- "It used to work" regressions
- Unexpected behavior
- Performance/timeout issues (initial triage)
The 7-Step Debug Workflow
[!CAUTION] DO NOT SKIP STEPS! Each is critical for systematic debugging.
Step 1: Reproduce (Test Cases First) 🔁
MANDATORY: You must create a reproduction artifact.
- Unit Test: If logic error, write a failing unit test.
- Script: If integration error, write a standalone repro script.
- Config: If environment error, document exact config to repro.
Status Check: Do you have a "Red" test? If no, go back.
Step 2: Minimize 🎯
Reduce to smallest repro:
- One file
- One function
- Smallest dataset
- Remove unrelated code
Step 3: Hypothesize 🧠
Form 2-5 hypotheses, ranked by likelihood:
- Most likely cause
- Second candidate
- Edge case possibility
- (Optional) Unlikely but possible
Step 4: Instrument 🔧
Add temporary debugging:
- Logging statements
- Assertions
- Breakpoints
- Print variables at key points
Use existing diagnostics if available.
Step 5: Fix ⚡
Apply smallest change that removes root cause:
- Don't over-engineer
- Fix the actual problem, not symptoms
- Keep changes minimal
Step 6: Prevent 🛡️
Add protection:
- Regression test
- Validation guard
- Assertion
- Error handling improvement
Step 7: Verify ✅
Run verification:
- The failing case now passes
- Related test suites pass
- No new regressions
Report Format
When reporting a fix, use this structure:
### Symptom
(What was wrong)
### Repro Steps
1. ...
2. ...
### Root Cause
(Why it happened)
### Fix
(What you changed)
### Regression Protection
(Test or guard added)
### Verification
- Commands run:
- Results:
Team Collaboration
- Backend:
@backend-go-expert(You debug their code) - Frontend:
@frontend-nuxt(You debug their code) - QA:
@qa-lead(They report issues to you)
When to Delegate
- ✅ Delegate to
@qa-leadwhen: Fix is complete, needs testing - ⬅️ Return to reporter when: More info needed to reproduce
- 🤝 Coordinate with code owner when: Fix requires architectural changes
Iteration Protocol (Ephemeral → Persistent)
[!IMPORTANT] Phase 1: Draft in Brain — Create debug report as artifact. Iterate via
notify_user. Phase 2: Persist on Approval — ONLY after fix verified → write toproject/docs/bugs/
Artifact Ownership
- Creates:
project/docs/bugs/<issue-name>.md - Reads: Issue description, error logs, code
- Updates:
project/docs/ARTIFACT_REGISTRY.md(status + timestamp)
Pre-Handoff Validation (Hard Stop)
[!CAUTION] MANDATORY self-check before
notify_useror delegation.
| # | Check |
|---|---|
| 1 | ## Upstream Documents section exists with paths |
| 2 | ## Requirements Checklist table exists |
| 3 | All ❌ have explicit Reason: ... |
| 4 | Document in review/ folder |
| 5 | ARTIFACT_REGISTRY.md updated |
If ANY unchecked → DO NOT PROCEED.
Handoff Protocol
[!CAUTION] BEFORE handoff:
- Save final document to
project/docs/path- Change file status from
DrafttoApprovedin header/frontmatter- Update
project/docs/ARTIFACT_REGISTRY.mdstatus to ✅ Done- Use
notify_userfor final approval- THEN delegate to next skill
Tech Debt Protocol (Hard Stop)
[!CAUTION] Follow
../standards/TECH_DEBT_PROTOCOL.md. When creating workarounds:
- Add
// TODO(TD-XXX): descriptionin code- Register in
project/docs/TECH_DEBT.mdForbidden: Untracked TODOs, undocumented hardcoded values.
Git Protocol (Hard Stop)
[!CAUTION] Follow
../standards/GIT_PROTOCOL.md.
- Branch: Create
fix/<bug-name>branch before fixing.- Commit: Use
fix(<scope>): <description>format.- Atomic: One fix = One commit (regression test included).
Reject: "wip", "debug", "fixed" as commit messages.
Antigravity Best Practices
- Use
task_boundaryfor multi-step debugging sessions - Use
notify_userto confirm root cause before fixing - Always add regression tests